Overview of models and methods for measuring economic
agent's expectations
Tine Janžek, Petra Ziherl
Abstract
Agents' expectations play an important role as one of the basic building blocks of theoretical
macroeconomic models. In practise, though, their intangible nature has always caused
consistency difficulties regarding their measurement and estimation methods. Increased
importance and use of econometric modelling in contemporary financial stability analyses
has triggered numerous advances in methods for measuring empirical expectations. Due to
the shortage of empirical data for a more detailed evaluation of expectation measures, this
article focuses more on the presentation of the most commonly established methods for their
capturing and the methodological aspect of the prevailing models.
KEYWORDS: expectations, rational expectations, surveys, qualitative measures, quantitative
measures, probabilistic measures, visual analog scale, probability approach, regression
approach, inflation, VAR.
Expectation examination approaches
Modern economic theory recognizes that the main difference between economics and
natural sciences lies in the forward-looking decisions made by economic agents.
Expectations play a key role in every segment of macroeconomics. In consumption theory
the paradigm life-cycle and permanent income approaches stress the role of expected future
incomes. In investment decisions present-value calculations are conditional on expected
future prices and sales. Asset prices (equity prices, interest rates, and exchange rates) also
clearly depend on expected future prices. Today we can distinct two major expectation
paradigms and concepts, where the second became the mainstream approach for further
development of expectation examination:
Adaptive expectations (AE) represent a hypothesis in economics which states that people
form their expectations about what will happen in the future based on what has happened in
the past. For example, if inflation has been higher than expected in the past, people would
revise expectations for the future. Adaptive expectations played a prominent role in
macroeconomics in the 1960s and 1970s. For example, inflation expectations were often
modelled adaptively in the analysis of the expectations-augmented Phillips curve. The
rational expectations revolution began with the observations that adaptive expectations, or
any other fixed-weight distributed lag formula, might provide poor forecasts in certain
contexts and that better forecast rules might be readily available.
Rational expectations (RE) state that agents' predictions of the future value of economically
relevant variables are not systematically wrong in the sense that all errors are random.
Equivalently, this is to say that agents' expectations equal true statistical expected values. An
1/9
alternative formulation is that rational expectations are model-consistent expectations, so the
agents inside the model assume the model's predictions are valid. In modern econometrics a
silent consensus has been established that at the theoretical level the rational expectations
hypothesis proposed by Muth (1961) has gained general acceptance as the dominant model
of expectations formation. His thesis provides a theory-consistent framework based on a
theory in which subjective expectations of individual decision makers are set to their
objective counterparts, assuming a known true underlying economic model.
1.
Development of expectation theory
Economic expectations are crucial in determining economic activity as they affect economic
decisions of consumers, politicians, businesses and economic experts. One of the first
records of how important the expectations are can be found in Keynes (1936), who
emphasised their role in the amount of output, employment and savings. He divided the
expectations into two types. First type is the short term expectations, which are concerned
with the price the producer expects to get for its product at the time he starts the production
process. The second, the long term expectations, are connected with what the entrepreneur
can hope to earn in the future if he purchases his products as an addition to his capital
equipment. The behaviour of each individual company is determined by its short term
expectations, which largely depend on the long term expectations of other parties. The
amount of employment in certain firm depends on these various expectations. However, a
change in expectations will only have full effect on employment over considerable period.
Samuelson (1948) showed that observation of consumption, price and income, combined
with basic assumption of consumer theory, implies restrictions on the consumption bundles
that this person would choose when having different budget sets with varying relative prices.
Muth (1961) argued that the economy generally does not waste information and that
expectations depend on the structure of the entire system. Expectation as informed
predictions of future events is the same as the predictions of the relevant economic theory.
According to his opinion dynamic economic models do not assume enough rationality.
Another author who contributes to the concept of rational expectations as one of the central
assumptions for many macroeconomic models is Lucas (1972). He studied the relation
between the rate of change in nominal prices and the level of real output, a variant of Phillips
curve. Demand shifts or monetary disturbances cause price movements. However, money
expansion has no real consequences in case all agents and their expectations behave
optimally. When placed in a setting in which the information conveyed to traders by market
prices is inadequate to permit them to distinguish real shifts from monetary fluctuation. In this
way monetary fluctuation leads to real output growth or fall.
McFadden (1974) showed that decisions made by people in random sample, combined with
assumptions on the population distribution of preferences, enable estimation of probabilistic
choice models. They can then be used to predict population choice behaviour in other
settings.
Kahneman and Tversky (1979, 1981) worked on the uncertainty phenomena under which
bounded rationality and expectations are formed, to explain divergences of economic
decision making from neo-classical theory. Their prospect theory, which they illustrate with
2/9
gambling and insurance business, distinguishes between an early phase of editing and a
subsequent phase of evaluation. The editing phase consists of the preliminary analysis of the
offered choices, while the former evaluates the prospects and chooses the one with the
highest value.
The function of the editing is to organize and reformulate one's options as to simplify the
subsequent evaluation choice. It can be affected by many things: one's attitudes towards
uncertainty and money, information or misinformation regarding probabilities and outcomes
and many others. In addition, people generally discard components that are shared by all
prospects under consideration, which leads to inconsistent preferences when the same
choice is presented in different forms. Gains and loses are defined relative to some neutral
reference point, usually one's current assets. However, there can be a discrepancy between
these two points, for example recent changes in wealth to which one has not yet adopted. A
change of reference point can alter the preference order for prospects. The frame is
controlled partly by formulation of the problem and partly by the norms, habits and personal
characteristics of the decision maker or even to a level of aspiration. The way the frame is
built represents a critical factor in the analysing of the decisions. Inconsistent responses to
the same problems can be framing effect with contradictory attitudes towards risks
connected with gains and losses. In other words, changing reference frame can determine
whether a certain outcome is evaluated as a gain or a loss.
Gigerenzer (1991) argues that Kahneman's and Tversky's results do not reflect respondents'
use of certain heuristic rules but are formed by the manner in which statistical information
was presented to them. Asking questions in terms of frequencies can reduce the magnitude
of the biases described by Kahneman and Tversky. Another argument is that their heuristics
are vaguely, atheoretically formulated, what limits their explanatory power as generators of
biases. Furthermore, according to Gigerenzer some of the biases characterized as errors
may be inappropriate. Nevertheless, Tversky and Kahmeman tried to explain formation of
expectation in real life and put doubt on the assumption of rational expectations.
Despite the prevail and domination of RE models in modern economics, an empirical
evidence confirmed that expectations formation is closely linked to point and density
forecasting and as such is subject to data and model uncertainty. Assuming that individuals
know the true model of the economy is no more credible than claiming that economic
forecasts made using econometric models will be free of systematic bias and informational
inefficiencies. This has led many investigators to explore the development of a weaker form
of the rational expectations hypothesis that allows for model uncertainty and learning.
Evans and Honkapohja (2001) presented the updated approach to expectation examination
in form of “adaptive learning”. It is a dynamic concept that combines RE with a certain
aspects of AE. The rational expectations approach supposes that economic agents have a
great deal of knowledge about the economy. In empirical work economists, who postulate
rational expectations, do not themselves know the parameter values and must estimate them
econometrically. It appears more natural to assume that the agents in the economy face the
same limitations on knowledge about the economy. So the concept of adaptive learning
assumes that agents adjust and revise their forecast (expectations) rules as new data
becomes available over time.
Manski (2004) concludes that most empirical research today concern choice problems in
which decision makers act with partial information. Common assumption is that persons form
3/9
probabilistic expectations for unknown quantities and maximizes expected utility, i.e. people's
expectations are objectively correct given the information they posses.
These recent findings together with contemporary adaptive learning models, where
experimental and survey data on expectations play an important role in providing better
insights into how expectations are formed, emphasized the role of the survey expectations.
2.
Survey expectations
Economists would ask people about their preferences and expectations when choice data
alone do not suffice to reveal the formation of decisions with partial information. The
collection of data on expectations of individual has its roots in the development of survey
methodology in the years before Second World War, when economists began to understand
the importance of future events on current decisions (Peseran and Weale, 2006). Use of
sample surveys made it possible to collect information way beyond what was covered by
administrative sources and full enumeration censuses.
The earliest systematic attempt to collect information on consumer expectations was the
study in 1944 conducted by Unites State Department of agriculture. It tried to measure
consumer sentiment. Currently the survey is run by the University of Michigan, with many
other similar surveys conducted across OECD countries. Key aspect is that some of its
questions have qualitative responses, e.g. respondents are simply asked whether they
expect to be better off or worse off.
Household surveys were later complemented with business survey on the state of economic
activity. Before Second World War countries did not have any formal indicators of sentiment.
One of the first surveys with questions about the expectations of output growth, price
increase in the near future, their recent movement and evolution of business environment
was done by Insitute für Wirtscaftsforschung in Munich in 1948. This survey structure has
since been adopted by other countries.
In 1946 Joseph Livingstone started a survey about expectations of inflation in which a
number of economists participated. Data collected were related with the macro-economy as
a whole. Another feature of the survey is that the respondents (as experts) were asked to
give point estimates of their expectation. Later the Livingston Survey has broadened in scope
to collect information on expectations about a range of economic variables. The survey is
now conducted by the Federal Reserve Bank of Philadelphia. Several similar surveys of
macroeconomic forecasts are also produced in USA.
However, economists were very sceptical of survey data at the beginning. In the 1950s and
early 1960s they even reported negative evidence on the usefulness in predicting consumer
purchase behaviour of expected household finances from such data. Juster (1966) proposes
that surveys of consumer intentions to buy provides inefficient predictors of purchase rates
as do not provide accurate estimates of mean purchase probability. Such surveys cannot
detect movements in mean probability among non-intenders. Purchase probabilities have
larger explanatory power of cross-section variance as their variable enables the formation of
subgroups with systematically different purchase rates.
4/9
Dominitz and Manski (1997) designed a survey specifically to elucidate information on
income uncertainty, and thereby produced an indication of subjective income uncertainty of
households. They concluded that the best way of collecting data on the subjective
distribution was to ask people about the probabilities that their incomes over the next twelve
months would be below each of four thresholds, with the thresholds chosen in the light of
reported current income. Respondents were also asked to report the lowest and highest
possible amounts their household incomes might be.
In the last 20 years the surveys has increasingly been used to research probabilistic
expectations of significant personal events (job loss, life expectancy, future income, etc). For
example, the Survey of Professional Forecasters asks the respondents to provide
probabilities within particular ranges, instead of point estimators.
3.
Expectation measures in surveys
To sum up, there general measurement methods of economic expectation have been
established over time (Stangl, 2008): qualitative measurement with category-rating scales
(usually three-category scale), quantitative measurement with point forecast (e.g. 2%
inflation rate) and measurement of economic expectation with subjective probabilities of
particular events.
Qualitative measures are quite popular since it is easier and quicker to give qualitative
response for respondents as they do not necessary need to consult their accounting records
and the wording is simpler for the phenomena which may be too complex to be described in
quantitative values. Therefore, such responses are less often a source of inaccuracies or
inconsistencies.
On the other hand, three category scales are very limited. Available options may not be
compatible with respondents' real opinions, which often gather answers towards central
category and thus imprecision and information loss. In addition, there is almost no
information on the dispersion of economic expectations, which is most commonly used as a
proxy of uncertainty. These problems can be to high extent solved by point forecasts or
quantitative information on outstanding orders, profits or turnover. However, because of the
confidentiality issues or time constraints, companies usually do not report it.
Pesaran and Weale (2006) suggest two conversion methods for converting the proportions of
qualitative answers into quantitative measures for the purpose of econometric expectations’
analysis. The main conversion techniques are:
1.
the probability approach of Carlson and Parkin (1975).
2.
the regression approach of Pesaran (1984) and Pesaran (1987)
Their method assumes that respondents have a common subjective probability distribution
over the future development of a variable and that they report a variable to go up or down if
the median of their subjective probability distribution lies above or below a threshold level.
The upper and lower thresholds which mark the so-called indifference limen are derived from
the respondents’ aggregate answers and the time-series properties of past realizations of the
macroeconomic variable under consideration. Most crucially, they assumed that the answers
are normally distributed with symmetric thresholds and they imposed that the average value
5/9
of past realizations and the average value of expectations must be equal, which is typically
referred to as the unbiasedness of expectations.
As an alternative to the Carlson-Parkin Method, Pesaran (1984) developed the Regression
Method. The basic idea is to use the relationship between realizations (measured by official
statistics) and respondents’ perceptions of the past (which is additionally queried in many
surveys) and to estimate the indifference limen on the basis of this observable data. In order
to quantify the respondents’ expectations about the future development of the variable under
consideration, Pesaran (1984) used these estimates and imposed them on the qualitative
expectations data. Thus, in contrast to the aforementioned methods, quantitative
expectations calculated by the Regression Method are a function of a specific regression
model, rather than a function of a specific probability distribution.
Despite the apparent advantages in using quantitative surveys, qualitative surveys are
widespread and considerable attention is paid to them in the real-time economic debate. It is
therefore important also to consider their performance as measures of the state of the macro
economy. There are three macroeconomic areas where qualitative surveys play the major
role for obtaining the data to assess the expectations: inflation, output growth and consumer
sentiment/consumer spending. The question of the link between expectational data and
inflation has received more attention than the one between expectational data and output
movements, partly because of the importance attached to inflationary expectations in a
number of macroeconomic models, such as the expectations-augmented Philips curve and
the assumption that a real interest rate can be derived by deducting inflationary expectations
from the nominal interest rate. The studies of consumer sentiment were the first surveys to
collect information on expectations. Dominitz and Manski (1997) provide a brief account of
early attempts to assess their value. They explain how the surveys acquired a poor
reputation because they seemed to have little capacity to explain future consumption.
Subjective probability distribution of future events is also an option in the survey. The results
enable to calculate the variance, which incorporate individual uncertainty and heterogeneity
of expectations. Mansky (2004) proposes another three advantages. First one is that
probability is a well-defined absolute numerical scale for responses that enables
interpersonally comparable responses. Furthermore, algebra of probability can be used to
examine the internal consistency about different events. Last but not least, subjective
probabilities can be compared with event frequencies, which allow the conclusion about the
correspondence between subjective beliefs and reality.
Nevertheless, probabilistic questions are time demanding and tend to cause high cognitive
load to respondents. Consequently, only these kinds of questions are applicable to only
people familiar with probability distributions. Another point for consideration is general
tendency of respondents to be optimistic, observed in variety of survey research.
Stangl (2008) proposes new measure of economic expectations. The visual analog scale
(VAS) is a qualitative measurement method, but it overcomes most of the problems with
traditional ones. Visual analog scales are rating scales on which a respondent ranks the
preferences along a continuous line or scale. They are easy to understand and to handle by
the respondents. VAS was first used in medical research as commonly used measures of
feeling and pain intensity.
6/9
Application of VAS in other surveys has been difficult so far due to costly operationalization.
With the burst of Internet surveys VAS has become easy to administer. Respondents can
express the direction of their attitude as well as the magnitude on a 1 to 100 point scale,
which put VAS close to interval scale measurement. By using graph scales it does not
increase cognitive load for the respondent and information collected is much broader. The
distributional shapes of responses and various measures of dispersion allow detect cyclical
turning-points earlier.
Stangl tests VAS for reliability for three reasons: the measure has not been used before, the
Internet and self-administration may negatively affect the scale reliability and for the reason
VAS responses overestimates people's discriminatory power relating to the subjects of
interests. She also argues that VAS business expectations contain two components:
heterogeneity of business expectations and uncertainty about the future economic
development. Overall, VAS is found to be reliable and valid information on economic
expectations.
Another of her findings is related to thresholds. They are asymmetrical around zero as the
respondents weight future losses stronger than gains. This means that expected positive
changes of economic situation have to be on considerably higher level before companies
report expected improvement in comparison with expected negative changes that would
make them report anticipated deterioration. This is also in line with Kahneman and Tversky
(1981).
Moreover, macroeconomic uncertainty broadens the indifference interval what leads to a shift
towards neutral category within the three category scale. According to Stangl (2008) results
indicate that the higher is the macroeconomic uncertainty the earlier respondents turn to
neutral category and the longer they remain within this state. The higher is the uncertainty,
the more extreme is the response to future losses compared to the response to future gains.
The respondents in Stangl case were experts; economists and executives in managerial
positions. VAS should therefore be further tested in general population to be accepted as the
measurement method of economic expectations.
4.
Methods for estimating expectations
Henzel and Wollmershäuser (2005) state that in empirical work expectations on future
macroeconomic variables can be treated in two ways: one is to set-up a theory on how
private agents form their expectations and the second way to introduce expectations into
empirical models is through direct measures of expectations derived from surveys of households, firms and other economic agents. The advantage of survey data is that expectations
are given exogenously in the context of a model, and that the nature of the expectations’
formation process can be investigated separately.
Kjellberg (2006) considers the following three methods as the most common for capturing
expectations:
•
futures method which utilizes financial market prices and derives market-based
expectations implicitly from prices of traded futures contracts;
7/9
•
the vector autoregression (VAR) forecast method which estimates a VAR-model and
uses the out-of-sample forecasts of the model to proxy expectations of a variable;
•
the survey method, which measures expectations by asking a sample of people
about their expectations of a variable.
In his study he thoroughly compared different approaches performing a systematic
comparative analysis of the methods where he tried to evaluate their suitability for the
purpose of economic forecasting. He estimated expectations using Federal funds rate target
(FFRT) and Federal funds rate (FFR). The FFRT is controller and used by the Federal
Reserve Bank (FED) to implement U.S. monetary policy and it is an important
macroeconomic and financial variable that is monitored by markets all over the world. FFR is
the traded market rate that the FED, by open market operations, keeps close to their stated
target rate (FFRT). They are typically very close.
His study revealed that despite of the fundamental difference among the methods used, the
survey measure and the futures measure turned out to be highly correlated (value of the
correlation coefficient is 0,81) which indicates that they manage to capture the same
phenomenon, that is true expectations. The survey method consistently overestimated the
realized changes in the interest rate. The VAR forecast method on the other hand showed
little resemblance with the other methods.
5.
Conclusion
Expectations and their measures remains an important subject, especially in contemporary
times of global financial crisis and instability of financial markets. The economic subjects
strive to reduce uncertainty by conducting different surveys and analyses, to produce reliable
forecasts that would enable more accurate foreseeable future and stabilize the economy.
Measuring and examination of expectations as a branch of economic science has received a
major attention in the last two decades, but remains an ongoing development. With
increased use of stochastic methods in economics and econometrics, estimation of
expectations overcame the corporate environment, where companies surveyed the market to
assess the demand for their products, and set itself as an important part of macroeconomics
and economic policies.
Theory of rational expectations has established as a preferred hypothesis that represents the
basis for the further development of the expectations' theory. Today the theory has evolved
with the inclusion of adaptive learning, where present expectations dynamically change by
implementation of additional available information. One of the approaches to examine this
phenomenon is by using vector autoregression models (VAR) with a certain amount of lag
within the impulses.
Besides technical econometric approaches modern authors have pointed out an increased
importance of survey measures. A lot of attention has been put towards different types of
surveys for different purposes: inflation forecasting, output forecasting, agents' sentiment
forecasting etc. In recent years a lot of efforts have been put towards quantification of
qualitative surveys and improving the qualitative measures. Future empirical evidence will
test the robustness and efficiency of the current efforts.
8/9
LITERATURE
1. Dominitz, J., & Manski, C. (1997). Using Expectations Data to Study Subjective
Income. Journal of the American Statistical Association 92, pp. 855–867.
2. Evans, G. W., & Honkapohja, S. (2001). Learning and Expectations in
Microeconomics. Princeton University Press.
3. Gigerenzer, G. (1991). Poznan Studies in the Philosophy of the Sciences and the
Humanities, Vol. 21. On Cognitive Illusions and Rationality, pp. 225-249.
4. Henzel, S., & Wollmershäuser, T. (2005, June). Ifo Working Paper No. 9. An
Alternative to the Carlson-Parkin Method for the Quantification of Qualitative Inflation
Expectations: Evidence from the Ifo World Economic Survey.
5. Juster, T. F. (1966). Consumer Buying Intentions and Purchase Probability: An
Experiment in Survey Design. Cambridge, Massachusetts: NBER Books, National
Bureau of Economic Research, Inc.
6. Kahneman, D., & Tversky, A. (1979, March). Prospect Theory: An Analysis of
Decision under Risk. Econometrica, Vol. 47, No. 2. , pp. 263-292.
7. Keynes, J. M. (1936). The General Theory of Employment, Interest and Money.
London: Macmillan Cambridge University Press.
8. Kjellberg, D. (2006). Measuring Expectations. In Working Paper Series 2006:9.
Uppsala: Uppsala University, Department of Economics.
9. Lucas, R. E. (1972). Journal of economic theory 4. Expectation and Neutrality of
Money, pp. 103-124.
10. Manski, C. F. (2004). Econometrica. Measuring expectations, pp. 1329-1376.
11. McFadden, D. (1974). Conditional logit anaysis of qualitative choice behaviour. In Z.
P., Frontiers in Econometrics (pp. 105-142). New York: Academic Press.
12. Muth, J. F. (1961, July). Econometrica, Vol. 29, No. 3. Rational Expectations and the
Theory of Price Movements, pp. 315-355.
13. Pesaran, M. (1984). Expectations formation and macroeconomic modelling. In P.
Magrange, & M. P., Contemporary Macroeconomic Modelling (pp. 27–53). Blackwell:
Oxford.
14. Pesaran, M. H., & Weale, M. (2006). Survey Expectations. In G. Elliott, C. Granger, &
A. Timmermann, Handbook of Economic Forecasting, Vol 1 (pp. 715-776). Elsevier.
15. Samuelson, P. A. (1948). Economics. McGraw-Hill Book Company, Inc.
16. Stangl, A. (2008). Essays on the Measurement of Economic Expectations. München:
Ludwig-Maximilians-Universität München.
17. Tversky, A., & Kahneman, D. (1981, January). The Framing of Decisions and the
Psychology of Choice. Science, New Series, Vol. 211, No. 4481 , pp. 453-458.
9/9